Temperature-field reconstruction and prediction are central to monitoring and assessing industrial and dynamic environments. A new arXiv preprint, GTSM, addresses this problem by aiming to recover four-dimensional spatio-temporal thermal fields.
The proposed method, named GTSM, combines Gaussian splatting with thermal state-space dynamics. The title suggests a hybrid approach: Gaussian splatting for spatial representation and state-space dynamics for temporal evolution, though the abstract excerpt does not detail the architecture.
The work is positioned for applications such as condition monitoring, process analysis, and state assessment. By targeting full 4D thermal fields, it could support better understanding of how temperature distributions change over time in complex settings.